An Assessment of Three Clinically Available Speech-in-Noise Test with Varying Contextual Cues

Submitted by:

Suzannah Boyle
University of South Florida

Presenter(s):

Suzannah Boyle, BS
Marsadi Parliament, AuD, PhD Student
Victoria Sanchez, AuD, PhD

Abstract

Objectives: Clinically, there is a lack of routine speech-in-noise testing due to a variety of barriers, including clinician perceived time constraints, uncertainty in which test material is most appropriate for a specific patient, and lack of materials which clearly quantify degrees of speech-in-noise impairment. The speech-in-noise tests available at present use a variety of target signals, such as digits, sentences, or monosyllabic words. Linguistic cues of speech, such as semantics, syntactics, word familiarity and word frequency, influence speech-in-noise performance; and, thus, may add to the clinician confusion about performance interpretation. Our study aimed to evaluate the performance between three speech-in-noise tests with varying linguistic cues currently available for clinical implementation: the Digits in Noise test (DIN), the Words in Noise test (WIN), and the American English Matrix Test (AEMT). To our knowledge, the AEMT, that is available in 15 languages, has yet to be compared to other clinically available speech-in-noise tests. We hypothesized that performance across tests would be correlated for both listeners with normal hearing and listeners with hearing loss. Understanding correlational relationships on these tests will both allow for standardization of degrees of speech-in-noise hearing loss and determine if these tests can be used interchangeably in clinic.

Design: Participants were 27 listeners with normal hearing and 32 listeners with sensorineural hearing loss. All participants were native English speakers. The outcome of interest was performance on the DIN (most linguistic cues), the AEMT (some linguistic cues), and the WIN (least amount of linguistic cues). The DIN and the WIN were presented in the presence of multi-talker babble and used a descending paradigm to derive the 50% correct recognition point using the Spearman-Karber Equation. The AEMT was presented in an open set in the presence of steady-state speech-spectrum noise and used an automated adaptive procedure to converge to the 50% correct recognition point.

Results: Speech-in-noise performance, reported as the signal to noise ratio (SNR) at which 50% correct word recognition is achieved (SNR-50), for the normal hearing listeners was the following: DIN: -12.6 dB SNR (SD = 1.94), AEMT: -7.8 dB SNR (SD = 4.22), and WIN: 4.8 dB SNR (SD = 1.65). Performance for the listeners with hearing loss was: DIN: -4.9 dB SNR (SD = 2.68), AEMT: -1.5 dB SNR (SD = 4.20), and WIN: 16.6 dB SNR (SD = 5.12). For both groups, performance increased with increasing linguistic cues. Strong positive relationships were found when assessing the materials using a Pearson product moment correlation with r values of: DIN/WIN: r= 0.849*, AEMT/WIN: r= 0.848*, and AEMT/DIN: r= 0.729* (*Indicates a p-value

Objectives

After this presentation, participants should be able to describe the characteristics of the Digits-In-Noise Test, the American English Matrix Test, and the Words In Noise Test.

After this presentation, participants should be able to compare and contrast the Digits-In-Noise Test, American English Matrix Test, and the Words In Noise Test.

After this presentation, participants should be able to describe speech-in-noise performance differences among different tests and identify which tests are most feasible to implement in a clinical setting.